Air pollution is a serious health risk, especially in middle-income countries. Monitoring is done by sensors. High quality sensors are available, but too expensive to be deployed in a dense network. The available low cost sensors have performance which is vulnerable to environmental conditions, and which varies between individual sensors.
Monitoring with low quality sensors lends itself to Bayesian modelling, which can be used to capture the measurement uncertainty and model the relationship between the sensor values and the true values. Colocating with existing high quality sensors can be used for calibration. Bayesian modelling can also be used to model pollution between sensors, and determine the best locations to deploy sensors to.